Working memory is a system for the temporary maintenance of information. It serves as the workbench of cognition, in that it provides access to the relevant information for the current task. Working memory has a severely limited capacity, and this capacity limit constrains our ability to deal with complex cognitive tasks such as reasoning, problem solving, planning, understanding difficult texts, or solving complicated equations.
One goal of the cognitive psychology of working memory is to characterize the limit of the system’s capacity. One class of theories states that the capacity is a fixed number of information elements (so-called “chunks”). Information given in a test of working memory either finds a free slot in the system, in which case it is retained with high accuracy, or else it is completely forgotten, and people can only guess about it. An alternative class of models assumes that every piece of information that people try to maintain in working memory is represented with a variable degree of strength. Failures of recall arise when the strength of an element falls below a threshold, or it is overshadowed by other, much stronger information. One goal of the project is to decide between these two views of working-memory capacity.
Much research on working-memory capacity is done with short-term recognition tasks, in which participants are asked to remember a small set of elements (e.g., words, or color patches) for a few seconds, and are then asked to decide for a further element whether or not it was part of the memorized set. We will therefore use short-term recognition as the main test bed for testing the relative merits of the two classes of theories. We will start by integrating three partial mathematical models of short-term recognition developed by the applicant. These models describe short-term recognition as driven by two processes: One is a quick and automatic assessment of the familiarity of the comparison stimulus, which reflects its summed similarity to all memory elements. The other is a slower process of recollection of individual memory elements, which are compared with the comparison stimulus. Dual-process theories can be regarded as an elaborate version of continuous-strength theories because at least the familiarity signal, and perhaps also the chance of successful recollection, is thought to depend on the continuous strength of memory elements.
Discrete-capacity theories of working memory derive support mostly from short-term recognition tasks with visual material (e.g., colors, shapes), whereas continuous-strength theories and dual-process theories gain support primarily from short-term recognition of verbal materials (e.g., words, letters). Our project will carry out parallel experiments with both kinds of experiments to allow a direct comparison of the merits of the competing models in both content domains. We will competitively test mathematical models assuming a discrete capacity, and models assuming continuous strength, including dual-task models, by fitting them to the data from our experiments.